Yeah you really don’t expect someone in their 20s playing at the highest level to be so dismissive. You can understand people who grew up before them but not Scheifele.
To me the issue is that the "analytics" community provides things that are measures or metrics but not truly predictive or prescriptive analytics. They truly are just metrics at this point.
For those that need the explanation a measure is just a count of things. ie. scored 1 goal. But knowing that the player scored 1 goal doesn't tell you much. If they had 1 shift, with 1 scoring chance, took 1 shot and scored 1 goal that's good. If they played 100 games and had 10 back door tap in attempts per game but only converted 1 time that's bad.
So by turning any measure into a percentage or ratio to help provide context you turn it into a metric. Goals per game, Goals per minute, Goals per shot attempt for example. Then you have a metric, but that alone doesn't really help you predict the future it only shows what happened in the past. You can try to do some trend analysis to see if something is an extreme outlier but mostly you're just comparing a player to their historical performance but it's missing a lot of context so correlation isn't necessarily causation.
What I mean by that is measures and metrics don't provide all the context. I.e. Corsi is just shot attempts but it doesn't tell you how dangerous that shot was, who took the shot from where, was the shooter pressured, was there a cross ice or low to high pass involved, was there traffic in front screening the goalie, did the shooter hit their target or not, how fast was the shot, was the goalie set when the shot came or not, was it a backhand shot, was the defender on the play left handed or right handed, where was the defenders stick, was the defender in the correct position or not, was the defender fatigued or not, was the attacking player fatigued or not, what was the score in the game, how much time was left in the game, who was the goalie, who were the defenders, who were the linemates, where were they positioned, etc.
All these things and more need to be considered when evaluating that shot that led to that 1 goal. Only by having accurately collected and accounted for all this information can you build a model that would show the likelihood of that 1 play resulting in a goal or not.
Given that most publicly available data models don't address or even contemplate these variables, how much stock should be put in the results if more than half the inputs in arriving at the result are missing? When shot location is registered, is it from where the player was or where the puck came off their stick?
So it's not that there's no value to the "Analytics" but they aren't nearly as predictive or prescriptive as people may like to have you think. What I mean by that is say you're driving towards a traffic light and it turns yellow. A predictive analytic would show you that the traffic light is yellow so it's about to turn red. Expected Goals doesn't necessarily tell you that the traffic light is yellow, it could still be green if all the variables were properly accounted for and just because it might be yellow doesn't mean that it's going to be red next.
Prescriptive analytics are the best and that would be that you're driving towards the traffic light, it has turned yellow so you need to stop the car because it's going to be red. They tell you what you should do based on the situation. i.e. should the player have shot that puck or not? Did shooting that puck help increase the odds of winning that game or not? With a prescriptive analytic you could show if the shot attempt was the optimal play or not and arrive at some sort of commentary about a players decision making and ability by looking at the results compared to what the model said the results should be.
I just don't think the data that is being consumed by the models is the quality that is needed to make any sort of definitive decision based on it. I think most analytics backers would tell you it's just 1 tool, not the only tool, because they also realize the limitations. Garbage in means garbage out.
(steps down off soapbox)